CN106940887A - A kind of satellite sequence image clouds of GF 4 and shadow detection method under cloud - Google Patents
A kind of satellite sequence image clouds of GF 4 and shadow detection method under cloud Download PDFInfo
- Publication number
- CN106940887A CN106940887A CN201710135417.2A CN201710135417A CN106940887A CN 106940887 A CN106940887 A CN 106940887A CN 201710135417 A CN201710135417 A CN 201710135417A CN 106940887 A CN106940887 A CN 106940887A
- Authority
- CN
- China
- Prior art keywords
- cloud
- image
- pixel
- sequence
- detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30192—Weather; Meteorology
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention is directed to No. four satellite images radiation pretreatment applications of high score, particularly cloud with shadow Detection under cloud using there is provided shadow detection method under a kind of No. four satellite sequence image clouds of high score and cloud.To the sequence image of same geographic area, Relative matching is realized by different linear functions, and by earth's surface average radiation brightness sequence, reduced by linear relative radiation normalizing and radiate difference caused by the acquisition time is different, according to the quantity of sequence image, select the algorithm filtered based on S G, or the automatic threshold method tag cloud based on statistics and shade under cloud, and the distance correction shade pixel testing result of the shade pixel and nearest cloud pixel according to detection.Committed step of the present invention is realized using ripe algorithm, with higher stability and applicability, and there is provided crucial technical support for the production of shadow Detection product and the lifting of Product Precision under the Yun Yuyun in being pre-processed for No. four satellite datas of high score.
Description
Technical field
The present invention relates to remote sensing images radiation treatment technology, specifically, it is related to a kind of for High Resolution Remote Sensing Satellites
Shadow Detection technology under the Yun Yuyun of sequence image.
Background technology
The radiation pretreatment of remote sensing images is always one of major subjects of Remote Sensing Data Processing.New remote sensing satellite input
After use, the problem of to the pretreatment of new data be critical.The pretreatment of remote sensing images generally comprises geometry and pre-processed in spoke
Pretreatment is penetrated, radiation pretreatment therein is in addition to crucial radiation calibration, and the statistics to image cloud amount is also one important
Step.According to the use demand of satellite data user side, cloud amount proportion turns into an important indicator of selection satellite image,
Such as tried one's best few image from cloud amount with many in drawing application in classification, and it is more more to pay close attention to cloud amount in the meteorological application with mitigation
Image.Overwhelming majority remote sensing satellite image data product all includes cloud amount information at present, is much all covered comprising special cloud
Wave band is marked, facilitates user to distinguish cloud and earth's surface pixel-by-pixel.The cloud detection method of optic of remote sensing satellite image is a lot, according to satellite data
The characteristics of the applicable cloud detection algorithm of exploitation be data prediction important step, such as wavelength band is with visible ray and near-infrared
Based on high-resolution multi-spectral remote sensing images cloud detection more use simple statistics with histogram combination automatic threshold method, or
Automatic threshold method of the person based on cloudless earth's surface reference data, this kind of method easily causes mistake for snowfield and highlighted dry earth's surface
Inspection；The low-temperature characteristics detection cloud of cloud layer, general essence are relied on observation satellite data comprising infrared band, that quantification degree is high more
Degree is higher, often comprising special cloud mark wave band in data product.
No. four satellites of high score (hereinafter referred to as GF-4) are that the geostationary orbit that China launches in December, 2015 is defended
Star, mounting space resolution ratio is the medium-wave infrared camera of 50 meters panchromatic, multispectral camera and 400 meters of resolution ratio, using face battle array
The mode of staring is imaged, and imaging interval possesses high time, the advantage of high spatial resolution soon to 20 seconds.From 2 months 2016 No. 3 states
Since anti-scientific and technological Industrial Development Bureau announces first batch of image, GF-4 has obtained China and neighboring area mass data, in detection forest
Played an important role in terms of fire, flood.The pretreatment of GF-4 satellite images equally includes geometry and radiation two
Part, geometry pretreatment includes structure, Ground control point matching and geometric exact correction of system imaging model etc., target be realize it is same
Imaging data is registering pixel-by-pixel under map projection.Radiation treatment includes detection of shade etc., target under radiation calibration, Yun Yuyun
It is so that image pixel value can accurately describe surface radiation situation.
The research and development of GF-4 satellite datas preconditioning technique will also on the basis of satellite data processing achievement before making full use of
Consider the characteristic of GF-4 satellite images in itself, research and develop special Processing Algorithm.According to the characteristics of GF-4 satellite datas, cloud detection is real
Existing technological approaches mainly has two：One is the low-temperature characteristics using cloud, according to cloud medium-wave infrared wave band brightness it is low with can
See the high Characteristics Detection of optical band brightness, this is to use more method at present；Two be the kinetic characteristic using cloud from sequence chart
Cloud is detected as in, because GF-4 satellites use geostationary orbit, and face battle array staring imaging mode, same geographic region is easily obtained
Substantial amounts of image under domain, according to the kinetic characteristic of cloud, shade under Yun Yuyun just can be distinguished using sequence image.
By the analysis and experiment to the specific data of GF-4, it is found that medium-wave infrared data are difficult to use in fine cloud detection,
Reason mainly has：Differences in resolution is excessive, and the medium-wave infrared pixel of 400 meters of resolution ratio corresponds to 8 × 8 block of pixels, is total to
The visible light pixel of 64 50 meters of resolution ratio；Covering geographic range misaligned and Pixel Dimensions are different, medium-wave infrared is 1204 ×
1024 pixels, it is seen that light is 10240 × 10240 pixels；Imaging time has differences, and medium-wave infrared is designed with visible images
For imaging time fixed intervals 45 seconds, the kinetic characteristic of cloud just make it that the position of both data medium clouds, form have differences；Data
In quality, also there is obvious noise problem in medium-wave infrared camera imaging.
The problem of for GF-4 satellite image cloud detection, in the case of medium-wave infrared image application difficult, research and utilization
The precision that sequence image improves cloud detection turns into a kind of effective technological approaches, and GF-4 is right in actual disaster monitoring application
The many days timing Continuous Observation data in disaster region generally have stronger sequence characteristic, this be based on several/sequence image
Cloud detection provides favourable condition.Engineering production GF-4 satellite data radiation prefinished products need a kind of sane sequence
Image cloud and shadow Detection algorithm under cloud.
The content of the invention
The purpose of the present invention is using there is provided under a kind of sequence image cloud and cloud for GF-4 fixed statellites image preprocessing
Shadow Detection technology, especially for panchromatic, multispectral image the L1 DBMS products of 50 meters of spatial resolutions of GF-4 satellites
In cloud detection production there is provided a kind of algorithm flow of shade wave band product under production marker cloud and cloud.This technology is based on
Ripe remote sensing images relative detector calibration algorithm and S-G (Savitzky-Golay) filtering algorithm, according to GF-4 satellite images
The quick cloud that customizes of radiation pretreatment demand and sequence image radiation characteristic and shadow Detection algorithm flow under cloud.
The present invention basic ideas be：The same geographic area sequential image data obtained for GF-4 fixed statellites is right
In the case of no system geometrical model, the linear function obtained using image Auto-matching realizes relative position between image
The registration pixel-by-pixel of relation, it is poor using radiation caused by different imaging times between automatic relative detector calibration reduction sequence image
It is different, correct earth's surface pixel-by-pixel in sequence image with reference to automatic threshold using S-G filtering, pass through the value before and after compared pixels amendment
Shade under Yun Yuyun is marked off, final output result is all corresponding single band cloud of each image and shade mark under cloud in sequence
Count evidence.
Described GF-4 satellite sequence images, are defined to panchromatic, the multispectral image of 50 meters of spatial resolutions, image four
The longitude and latitude difference of angle point is no more than ± 0.3 degree, can be GF-4 satellites stare the sequence image that is obtained under pattern or
Do not obtain on the same day, according to the sequence image for obtaining time order and function arrangement.
The GF-4 satellite sequence image clouds that technical scheme is provided and shadow detection method under cloud, it is characterised in that
Including following implementation steps：
A data predictions, obtain sequence image Relative matching linear dimensions；
The linear relative radiation normalizings of B, automatically extract sequence image pseudo- invariant features culture point between any two, are compared by counting
The pseudo- invariant features culture point radiation difference of more each image, finds out the big data of integral radiation difference and carries out relative detector calibration；
C sequence image cloud detection, according to the quantity of sequence image, selects the algorithm filtered based on S-G, or based on statistics
Automatic threshold method tag cloud and cloud under shade, obtain shadow mask wave band data under the Yun Yuyun of each image；
D corrects testing result, first according to the distance correction shade pixel of the shade pixel of detection and nearest cloud pixel, then
Show that artwork finds out testing result data of problems with cloud detection result figure and carries out cloud sector amendment by laminated structure, if
In the presence of the data of several cloud detection results difference, then these data are constituted into new sequence image and re-started at above-mentioned cloud detection
Reason.
Above-mentioned implementation steps are characterised by：
Data prediction described in step A, including data integrity inspection, geographical covering are carried out to the sequence image of input
Range check, some preparation initialization process by earth's surface average radiation brightness sequence and program operation.The acquisition sequence
Image Relative matching linear dimensions, detailed process is the sequence image to same geographic area, utilizes the near of four angle points of image
The substantially relative position relation between image is determined like latitude and longitude coordinates, and image block Auto-matching is determined according to position error
Range of search, by Auto-matching, obtains and controls point data and to fit pass through difference between a linear function, sequence image
Linear function realize Relative matching, without entering the conversion such as row interpolation and resampling in itself to image.
Linear relative radiation normalizing described in step B, for reducing sequence image, factor data obtains the time between any two
Difference caused by radiation differences, the relative detector calibration technology in remote sensing fields is used in sequence between adjacent two images,
Here IR-MAD (the re-weighted Multivariate Alteration based on multi-source canonical correlation analysis are used
Detection transformation) change the pseudo- invariant features culture point realized and automatically extract two images；It is described to pass through system
The pseudo- invariant features culture point radiation difference of more each image of meter, the comparative approach of use be for the piece image in sequence, its
Size of the average difference in whole sequence between the pseudo- invariant features culture point of the image zooming-out before and after sequence is compared, such as
Really in the figure and sequence before and after pseudo- constant culture point average difference between image it is all very big, then need to carry out the image relative
Radiant correction；The big data of integral radiation difference of finding out carry out relative detector calibration, and relative detector calibration is using pseudo- constant
The linear function that culture point is fitted is completed.
Sequence image cloud detection described in step C, uses different detection methods according to the quantity of sequence image, works as image
Quantity is more than or equal to 10, selects the algorithm filtered based on S-G, and sequential filtering pixel-by-pixel is first carried out to sequence image, further according to
Cloud and shade under cloud are distinguished in the comparison of pixel value before and after each image filtering；When amount of images is less than 10, then direct statistical series
The average and intermediate value of image pixel by pixel, according under the diff area cloud and cloud of the average and intermediate value of each image pixel value and statistics
Shade；Shadow mask wave band data under the Yun Yuyun of each image of step acquisition；
Testing result is corrected described in step D, first according to the distance correction of the shade pixel of detection and nearest cloud pixel
Shade pixel, according to the ultimate range of shadow distance cloud under cloud, such as GF-4 images can the prime number of value 500 or 1000, for
The shade pixel each detected, if not finding the cloud pixel of detection in the radius pixel coverage, is determined as flase drop,
The pixel is rejected from shade pixel；Show that artwork and cloud detection result figure are found out testing result presence and asked by laminated structure
The data progress cloud sector amendment of topic, data of problems described here, mainly for two kinds of situations, one is that testing result is present
More fragment and leak, now remove fragment using the morphologic method of computer and fill leak, and two be that testing result is present
The obvious flase drop of vast scale, such as be cloud by flase drops such as highlighted earth's surface, the waters surface, now need detected again to change plan, such as
There are the data of several cloud detection results difference in fruit, then these data are constituted into new sequence image re-starts Cloud detection.
The present invention has following features compared with prior art：GF-4 sequence images cloud and cloud are directed to the invention provides one kind
Lower shadow Detection solution, the rapid registering of GF-4 sequence images position relationship pixel-by-pixel is realized by linear function, is utilized
Linear relative detector calibration reduces the radiation difference between data, and shade under Yun Yuyun is marked off using S-G filter results.Algorithm is certainly
Dynamicization degree is high, and shadow Detection process is without man-machine interaction under Yun Yuyun, and user only needs to carry out simply final testing result
Inspection, individual data is handled again.The committed step being related to is realized using ripe algorithm, with higher stabilization
Property and applicability.The production of shadow Detection product and carrying for Product Precision under Yun Yuyun in being pre-processed for GF-4 satellite datas
There is provided crucial technical support for liter.
Brief description of the drawings：
Fig. 1 is GF-4 satellite sequence image cloud detection flow charts
Fig. 2 is the schematic diagram of S-G filtering and threshold value cloud detection at single pixel
Fig. 3 is testing result morphology amendment schematic diagram
Embodiment：
The thought of this technology is to realize cloud and shade under cloud using kinetic characteristic of the shade under Yun Yuyun in sequence image
Detection, its necessary condition is：The data that GF-4 satellites are obtained easily constitute sequence, and the same area obtained in different time
The position relationship of rotation and translation is only existed between multiple image.The necessary condition is to meet for GF-4 satellite images, is
Because GF-4 satellites use geostationary-satellite orbit, the position relative to the earth is fixed, and its imaging geometry is constant, including
Geometrical relationship to any point in the range of earth Observable to satellite sensor imaging point is all fixed.Fixed statellite position
The stationarity put, it is ensured that to multiple observed images of the same area, in the picture heart point and four angular coordinate identical feelings
Under condition, the system imaging geometrical model of all images is identical so that the systematical distortion and spatial resolution of image are all one
Cause.And GF-4 satellites are imaged using the face battle array mode of staring, and imaging time is instantaneous, and the imaging process of image will not be introduced
New geometric distortion, the shake of satellite and the shake of sensor are all difficult that imaging is impacted.Finally, GF-4 satellite imageries are determined
Position high precision.By pointing to control, GF-4 can be realized carries out free observation to China and surrounding area, can also use and stare pattern
FX to the km of breadth 400 is persistently observed, and its positioning precision reaches ± 0.1 degree so that GF-4 satellites, which have, to be obtained
The ability of the sequence image of same FX.Since first batch of image being announced from 2 months 2016 No. 3 defense-related science, technology and industry offices, GF-
4 satellites have obtained China and neighboring area mass data, wherein comprising the substantial amounts of data that may make up sequence, including GF-4
Satellite, which is stared, can obtain the sequence image at close moment on the same day under pattern, also there is the same area image construction not obtained on the same day
Sequence.
Realize GF-4 satellite sequence image clouds with shadow Detection flow under cloud as shown in figure 1, in conjunction with attached using the present invention
Figure is described.
The data prediction of processing unit 111, data prediction is directed to GF-4 satellite sequence images, for GF-4 data publications
50 meters of spatial resolutions visible ray and near-infrared image, there is two ways to obtain the image of sequence：One is GF-4 phases
The sequence for the continuous multiple image obtained on the same day, in similar time the composition that machine is obtained in the case where staring mode of operation；Two right and wrong
Stare under pattern, multiple image do not obtain on the same day, areal.In the actual operation of GF-4 satellites, obtain second
The non-sequential image data for staring pattern composition is easy to.Because disaster monitors the vital task as GF-4, when a certain region
During generation disaster, GF-4 can carry out many days, repeatedly observation in the range of 0.1 degree of longitude and latitude error to FX, so that
Constitute sequence image.Preprocessing algorithms program carries out data integrity inspection, geographical covering model to the sequence image of input
Enclose some preparation initialization process of inspection, sequence permutation and program operation.Wherein sequence permutation is not to be obtained according to data
The time taken is ranked up, but is sorted according to the average radiation brightness of each image earth's surface.
According to the average radiation brightness of each image earth's surface sequence in the present invention, for the follow-up of GF-4 fixed statellite data
Processing is crucial.It is close, ground at the time of geographical position image same different from the satellite in Sun-synchronous orbit acquisition earth
Ball geosynchronous satellite is constant with respect to position of the earth, any time imaging in may be selected one day, is not imaged in the same time in daytime
According to the difference of altitude of the sun, image integral radiation brightness is also different, such as 8 o'clock of morning is imaged with 12 o'clock of high noon
Image radiation difference.Specific sort method is the histogram of blue wave band in statistical series image, is excluded bright with crossing dark pixel
After value, using the average of residual pixel as sequence foundation.Here it is that filter out can to exclude the bright purposes with crossing dark pixel values
Shade under the Yun Yuyun of energy, it is ensured that obtained average, which is tried one's best, represents the radiance situation of earth's surface.
After sequence image sequence, due to being pre-processed the invention belongs to the radiation of GF-4 data, the image of use is without system
The initial data of geometric correction, the registering pixel-by-pixel of atural object can not be realized according only to the directly superposition of framing information.Here adopt
With linear Relative matching method, detailed process is the sequence image to same geographic area, utilizes the approximate of four angle points of image
Latitude and longitude coordinates determine the substantially relative position relation between image, and determine according to position error the inspection of image block Auto-matching
Rope scope, by Auto-matching, obtains control point data and fits logical between a linear function y=ax+b, sequence image
Cross different linear function parameter a and b and realize Relative matching, without entering the conversion such as row interpolation and resampling in itself to image.
By the sequential image data of pretreatment, if being the non-different number of days evidence for staring pattern acquiring, need to carry out
The linear relative radiation normalization of processing unit 112.The processing is the committed step of algorithm flow, for reducing sequence image
Factor data obtains in the caused radiation difference of difference of time, sequence and uses remote sensing fields between adjacent two images between any two
In relative detector calibration technology, for the piece image in sequence, its pseudo- invariant features with the image zooming-out before and after sequence
Size of the average difference in whole sequence between culture point compares, if the puppet before and after in the figure and sequence between image is not
Become culture point average difference all very big, then need to carry out relative detector calibration to the image.Here changed using IR-MAD and realized
Automatically extract the pseudo- invariant features culture point of two images.
IR-MAD conversion comes from the MAD conversion of Nielsen et al. (1998) propositions, and the algorithm is in order to cover two phases
Change pixel in image, is initially formed the linear combination of pixel value in the N number of passage of two images.Distinguished with random vector X and Y
Represent target figure and with reference to the pixel value filtered out in figure overlay region.According to following transformation for mula：
U=aTX=a1X1+a2X2+Λ+aNXN
V=bTY=b1Y1+b2Y2+Λ+bNYN
Wherein aiWith biFor MAD coefficients, MAD conversion minimizes the positive correlation between U and V.Submitting to restraint：Var (U)=
On the premise of Var (V)=1, MAD variables are defined：
MAD=Var (U-V)=Var (U)+Var (V) -2cov (U, V)=2 (1-corr (U, V)) → Maximum
Minimize the statistic processes that positive correlation coefficient corr (U, V) is a standard, i.e., so-called generalized eigenvalue problem.
MAD variables each component obtained is mutually orthogonal, and is the invariant of linear transformation.Why the present invention selects MAD to convert
To extract invariant features point, this characteristic insensitive to the linear relationship between variable X and Y converted just because of MAD can
To be well adapted for the relatively large radiation difference existed between the GF-4 images of different time acquisition.IR-MAD conversion is further to improve
The precision and stability of MAD algorithms.
Change the pseudo- invariant features culture point of two images automatically extracted out using IR-MAD, intended using least square method
The linear function y=ax+b of an entirety is closed out, will be radiated using traditional linear relative detector calibration of remote sensing images in sequence
Radiation level of the big image rectification of difference to adjacent piece image.
Whether the picture number included according to sequence image is more than 10 width, determines that subsequent treatment uses processing unit 113, or
Person's processing unit 114.
Processing unit 113 is counted is with automatic threshold cloud detection, specific algorithm, if sequence image includes n width images, n
≤ 10, to the location of pixels of each Relative matching in sequence image, the average Vmin and intermediate value Vmid of n pixel are counted,
If average is numerically more or less the same with intermediate value, such as | Vmin-Vmid | ＜ 10, then all mark is n pixel.If
Intermediate value is big with average numerical value difference, then is gradually compared n pixel with Vmid, if Vi-Vmid ＞ Vcloud, judge
Ith pixel is cloud, and Vcloud is cloud threshold value；If Vi-Vmid ＜ Vshadow, shade under judging ith pixel as cloud,
Vshadow is shadow thresholds under cloud, is negative value；Vcloud's and Vshadow can value ± 2 | Vmin-Vmid |, or ± 3 |
Vmin-Vmid|。
Processing unit 114S-G is filtered and threshold value cloud detection, and step processing is first to carry out S-G filtering to sequence image, then is led to
Cross the variation of numerical value before and after comparing filtering and determine whether shade under Yun Yuyun.
S-G filtering is reached by sliding window fitting of a polynomial carries out smooth purpose (Savitzky＆ to sequence data
Golay, 1964).Sequence number is N, and carrying out the fitting of k (k≤n) rank multinomial to wherein length for n=2m+1 subsequence can table
It is shown as：
S-G filterings are to the certain point t in sequence0And its common n=2m+1 point (ti, yi) of left and right m neighborhoods, i ∈
[- m, m], carries out the fitting of a polynomial of k ranks (k≤n), with the data (t at the sliding window center after fitting0, y0) displacement it is original when
Between data (t in sequence0, y0), then move right window, window center is moved to next data in sequence, repeats above-mentioned mistake
Journey, until sliding window reaches sequence end.Smooth window coefficient is tried to achieve by least square method mode.
If sequence image include n width images, n ＞ 10, to the location of pixels of each Relative matching in sequence image,
After S-G is filtered, by n pixel ViGradually with filtered value Vi-SGIt is compared, if Vi-Vi-SG＞ Vcloud, then judge
Ith pixel is cloud；If Vi-Vi-SG＜ Vshadow, then shade under judging ith pixel as cloud；VcloudWith VshadowIt is desirable
Empirical value, such as 20 or 30, according to specific data cases adjustable thresholds.In sequence image at single pixel S-G filtering with
The schematic diagram of threshold value cloud detection is shown in Fig. 2.
Testing result to shade under cloud is modified, and is repaiied according to the distance of the shade pixel of detection and nearest cloud pixel
Just, provide the ultimate range of shadow distance cloud under cloud, such as GF-4 images can the prime number of value 500 or 1000, for each inspection
The shade pixel measured, if not finding the cloud pixel of detection in the radius pixel coverage, is determined as flase drop, by the picture
Member is rejected from shade pixel.
The precision suggestion of testing result carries out hand inspection, can be logical according to specific data cases for the poor result of precision
Handled again after crossing adjustment threshold value.It is based on single pixel, it sometimes appear that testing result office additionally, due to cloud detection
There is the situation of a large amount of fragments and leak in portion region.In order to improve the boundary effect of cloud detection, using the morphologic side of computer
Method carries out cloud sector ornamenting processing, removes in the isolated cloud sector for being less than certain pixel count outside cloud border, filling cloud border less than certain
The cavity of pixel count, then ornamenting cloud border.Effect diagram is shown in Fig. 3.Because actual cloud is also likely to be too discrete in itself
, determined so whether carrying out ornamenting to cloud sector by user.
The result of shadow Detection saves as shade value 1, cloud under 8 single band images, earth's surface value 0, cloud under Yun Yuyun
Value 2.N width sequence image correspondence n width testing results, user is supplied to as GF-4 primary data products.
The example of the present invention is realized on a pc platform, and user side has been delivered at present and has been tested and is used, as
GF-4 data radiation pretreatment medium cloud characteristic parameter inverting key technology.
It should be pointed out that the above embodiment can make those skilled in the art that this hair is more fully understood
It is bright, but do not limit the invention in any way.Therefore, it will be appreciated by those skilled in the art that still can be to present invention progress
Modification or equivalent；And technical scheme and its improvement of all spirit and technical spirit that do not depart from the present invention, it all should
Cover among the protection domain of patent of the present invention.
Claims (4)
1. a kind of GF-4 satellite sequences image cloud and shadow detection method under cloud, this method is for No. four satellite image radiation of high score
Pretreatment application, particularly cloud and shadow Detection application under cloud, it is characterised in that including following implementation steps：
A data predictions, obtain sequence image Relative matching linear dimensions；It is bright by earth's surface average radiation in the data prediction
Degree sequence；Sequence image Relative matching linear dimensions is obtained, detailed process is the sequence image to same geographic area, utilizes figure
As the approximate latitude and longitude coordinates of four angle points determine the substantially relative position relation between image, and image is determined according to position error
The range of search of piecemeal Auto-matching, by Auto-matching, obtains control point data and fits a linear function, sequence chart
Relative matching is realized by different linear functions as between, without entering the conversion such as row interpolation and resampling in itself to image；
The linear relative radiation normalizings of B, automatically extract sequence image pseudo- invariant features culture point between any two, each by statistical comparison
Image puppet invariant features culture point radiation difference, finds out the big data of integral radiation difference and carries out relative detector calibration；The line
Property relative radiation normalizing reduce sequence image factor data obtain phase in radiation differences caused by the difference of time, sequence between any two
The relative detector calibration technology in remote sensing fields is used between adjacent two images；By the pseudo- invariant features of each image of statistical comparison
Object point radiates difference, and the comparative approach of use is its puppet with the image zooming-out before and after sequence for the piece image in sequence
Size of the average difference in whole sequence between invariant features culture point compares, if before and after in the figure and sequence image it
Between pseudo- constant culture point average difference it is all very big, then need to carry out relative detector calibration to the image.
C sequence image cloud detection, according to the quantity of sequence image, selects the algorithm filtered based on S-G, or based on statistics from
Dynamic threshold method tag cloud and shade under cloud, obtain shadow mask wave band data under the Yun Yuyun of each image.
2. the method according to claim 1, it is characterised in that：
Sequence image cloud detection, according to the quantity of sequence image use different detection methods, when amount of images be more than or equal to 10,
The algorithm filtered based on S-G is selected, sequential filtering pixel-by-pixel is first carried out to sequence image, before and after each image filtering
Cloud and shade under cloud are distinguished in the comparison of pixel value；When amount of images is less than 10, the then average of direct statistical series image pixel by pixel
With intermediate value, according to each image pixel value and shade under the average of statistics and diff area cloud and the cloud of intermediate value.
3. the method according to claim 1, it is characterised in that：
Statistics and automatic threshold cloud detection, specific algorithm is, if sequence image includes n width images, n≤10, to sequence image
In each Relative matching location of pixels, count n pixel average Vmin and intermediate value Vmid, if average and median numbers
Be more or less the same in value, such as | Vmin-Vmid | ＜ 10, then all mark is n pixel；If intermediate value and average numerical difference
It is different big, then n pixel is gradually compared with Vmid, if Vi-Vmid ＞ Vcloud, judges ith pixel as cloud,
Vcloud is cloud threshold value；If Vi-Vmid ＜ Vshadow, shade under judging ith pixel as cloud, Vshadow is Yun Xiayin
Shadow threshold value, is negative value；Vcloud's and Vshadow can value ± 2 | Vmin-Vmid |, or ± 3 | Vmin-Vmid |.
4. the method according to claim 1, it is characterised in that：
Testing result is corrected, according to the shade pixel of detection and the distance correction shade pixel of nearest cloud pixel, according under cloud
The ultimate range of shadow distance cloud, such as GF-4 images can the prime number of value 500 or 1000, for each direct-shadow image detected
Member, if not finding the cloud pixel of detection in the radius pixel coverage, is determined as flase drop, by the pixel from shade pixel
It is middle to reject.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710135417.2A CN106940887B (en) | 2017-03-09 | 2017-03-09 | GF-4 satellite sequence image cloud and cloud shadow detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710135417.2A CN106940887B (en) | 2017-03-09 | 2017-03-09 | GF-4 satellite sequence image cloud and cloud shadow detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106940887A true CN106940887A (en) | 2017-07-11 |
| CN106940887B CN106940887B (en) | 2020-04-14 |
Family
ID=59469719
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710135417.2A Active CN106940887B (en) | 2017-03-09 | 2017-03-09 | GF-4 satellite sequence image cloud and cloud shadow detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106940887B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108051371A (en) * | 2017-12-01 | 2018-05-18 | 河北省科学院地理科学研究所 | A kind of shadow extraction method of ecology-oriented environment parameter remote-sensing inversion |
| CN109671038A (en) * | 2018-12-27 | 2019-04-23 | 哈尔滨工业大学 | One kind is based on the classified and layered relative radiometric correction method of pseudo- invariant features point |
| CN109918523A (en) * | 2019-02-14 | 2019-06-21 | 广东工业大学 | A kind of circuit board element detection method based on YOLO9000 algorithm |
| CN110070513A (en) * | 2019-04-30 | 2019-07-30 | 上海同繁勘测工程科技有限公司 | The radiation correction method and system of remote sensing image |
| CN110555818A (en) * | 2019-09-09 | 2019-12-10 | 中国科学院遥感与数字地球研究所 | method and device for repairing cloud region of satellite image sequence |
| CN111709458A (en) * | 2020-05-25 | 2020-09-25 | 中国自然资源航空物探遥感中心 | Automatic quality inspection method for top-resolution five-number images |
| CN116934745A (en) * | 2023-09-14 | 2023-10-24 | 创新奇智(浙江)科技有限公司 | Quality detection method and detection system for electronic component plugging clip |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102637301A (en) * | 2012-03-20 | 2012-08-15 | 武汉大学 | Method for automatically evaluating color quality of image during aerial photography in real time |
| CN102750701A (en) * | 2012-06-15 | 2012-10-24 | 西安电子科技大学 | Method for detecting spissatus and spissatus shadow based on Landsat thematic mapper (TM) images and Landsat enhanced thematic mapper (ETM) images |
| CN103383773A (en) * | 2013-03-26 | 2013-11-06 | 中国科学院遥感与数字地球研究所 | Automatic ortho-rectification frame and method for dynamically extracting remote sensing satellite image of image control points |
-
2017
- 2017-03-09 CN CN201710135417.2A patent/CN106940887B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102637301A (en) * | 2012-03-20 | 2012-08-15 | 武汉大学 | Method for automatically evaluating color quality of image during aerial photography in real time |
| CN102750701A (en) * | 2012-06-15 | 2012-10-24 | 西安电子科技大学 | Method for detecting spissatus and spissatus shadow based on Landsat thematic mapper (TM) images and Landsat enhanced thematic mapper (ETM) images |
| CN103383773A (en) * | 2013-03-26 | 2013-11-06 | 中国科学院遥感与数字地球研究所 | Automatic ortho-rectification frame and method for dynamically extracting remote sensing satellite image of image control points |
Non-Patent Citations (2)
| Title |
|---|
| 卢晶等: "基于风云3C卫星双氧通道的云检测算法", 《科学技术与工程》 * |
| 谢俊德: "landsat遥感图像云检测及薄云去除的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108051371B (en) * | 2017-12-01 | 2018-10-02 | 河北省科学院地理科学研究所 | A kind of shadow extraction method of ecology-oriented environment parameter remote-sensing inversion |
| CN108051371A (en) * | 2017-12-01 | 2018-05-18 | 河北省科学院地理科学研究所 | A kind of shadow extraction method of ecology-oriented environment parameter remote-sensing inversion |
| CN109671038A (en) * | 2018-12-27 | 2019-04-23 | 哈尔滨工业大学 | One kind is based on the classified and layered relative radiometric correction method of pseudo- invariant features point |
| CN109671038B (en) * | 2018-12-27 | 2023-04-28 | 哈尔滨工业大学 | Relative radiation correction method based on pseudo-invariant feature point classification layering |
| CN109918523A (en) * | 2019-02-14 | 2019-06-21 | 广东工业大学 | A kind of circuit board element detection method based on YOLO9000 algorithm |
| CN109918523B (en) * | 2019-02-14 | 2021-03-30 | 广东工业大学 | Circuit board component detection method based on YOLO9000 algorithm |
| CN110070513B (en) * | 2019-04-30 | 2021-10-01 | 上海同繁勘测工程科技有限公司 | Radiation correction method and system for remote sensing image |
| CN110070513A (en) * | 2019-04-30 | 2019-07-30 | 上海同繁勘测工程科技有限公司 | The radiation correction method and system of remote sensing image |
| CN110555818B (en) * | 2019-09-09 | 2022-02-18 | 中国科学院遥感与数字地球研究所 | Method and device for repairing cloud region of satellite image sequence |
| CN110555818A (en) * | 2019-09-09 | 2019-12-10 | 中国科学院遥感与数字地球研究所 | method and device for repairing cloud region of satellite image sequence |
| CN111709458A (en) * | 2020-05-25 | 2020-09-25 | 中国自然资源航空物探遥感中心 | Automatic quality inspection method for top-resolution five-number images |
| CN116934745A (en) * | 2023-09-14 | 2023-10-24 | 创新奇智(浙江)科技有限公司 | Quality detection method and detection system for electronic component plugging clip |
| CN116934745B (en) * | 2023-09-14 | 2023-12-19 | 创新奇智(浙江)科技有限公司 | Quality detection method and detection system for electronic component plugging clip |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106940887B (en) | 2020-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106940887A (en) | A kind of satellite sequence image clouds of GF 4 and shadow detection method under cloud | |
| Fisher et al. | Comparing Landsat water index methods for automated water classification in eastern Australia | |
| Ashby et al. | Seds: The spitzer extended deep survey. survey design, photometry, and deep irac source counts | |
| CN110186820A (en) | Multisource data fusion and environomental pollution source and pollutant distribution analysis method | |
| Roy et al. | Multi-temporal MODIS–Landsat data fusion for relative radiometric normalization, gap filling, and prediction of Landsat data | |
| Pirzkal et al. | FIGS—faint infrared grism survey: description and data reduction | |
| CN102565778B (en) | Relative radiometric correction method for automatically extracting pseudo-invariant features for remote sensing image | |
| CN111368817B (en) | Method and system for quantitatively evaluating thermal effect based on earth surface type | |
| Feng et al. | Land adjacency effects on MODIS A qua top‐of‐atmosphere radiance in the shortwave infrared: S tatistical assessment and correction | |
| CN113537018B (en) | Water and soil conservation monitoring method based on multi-time-phase satellite remote sensing and unmanned aerial vehicle technology | |
| Everett et al. | High-resolution multi-band imaging for validation and characterization of small Kepler planets | |
| CN109376600A (en) | Multi-spectrum remote sensing image comprehensive characteristics cloud detection method of optic and device | |
| Gobron et al. | Evaluation of the MERIS/ENVISAT FAPAR product | |
| Bruce et al. | Pre-processing methodology for application to Landsat TM/ETM+ imagery of the wet tropics | |
| Liu et al. | An improved land target-based atmospheric correction method for Lake Taihu | |
| CN114022783A (en) | Satellite image-based water and soil conservation ecological function remote sensing monitoring method and device | |
| CN106887016A (en) | A kind of automatic Relative matching method of the satellite sequence images of GF 4 | |
| Spencer et al. | Exploring aerosols near clouds with high‐spatial‐resolution aircraft remote sensing during SEAC4RS | |
| Wang et al. | An adaptive atmospheric correction algorithm for the effective adjacency effect correction of submeter-scale spatial resolution optical satellite images: Application to a WorldView-3 panchromatic image | |
| CN106950551B (en) | SAR radiometric calibration method based on radiometric stability characteristic mining | |
| CN115187481A (en) | Airborne push-broom hyperspectral image radiation disturbance correction method | |
| CN110702228B (en) | Edge radiation correction method for aviation hyperspectral image | |
| Aleksanin et al. | Specificity of atmospheric correction of satellite data on ocean color in the Far East | |
| Han et al. | Hypothetical visible bands of advanced meteorological imager onboard the geostationary Korea multi-purpose satellite-2A using data-to-data translation | |
| CN111199557A (en) | Quantitative analysis method and system for decay of remote sensor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |